Loading...

Personalized Diverity in Recommender Systems

Mehrjoo, Mehrdad | 2015

483 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 47421 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Jalili, Mahdi
  7. Abstract:
  8. Recommender systems are in the center of network science and are becoming increasingly important in individual business for providing efficient personalized services and products to users. The focus of previous research in the field of recommendation systems were on improving the accuracy of the system through designing more accurate recommendation lists. Recently, the community has been paying attention to diversity and novelty of recommendation list as key characteristics of modern recommender systems. In many cases, novelty and precision do not go in the same direction and the accuracy-novelty dilemma is one of the challenging problems in recommender systems, which needs efforts in making a trade-off between them. In this poject, we introduce a new metric for recommender systems which is related to interest rate of users to diversity in their choices. It is obvious that, each two people have different interest about how diverse items they like to use, or how diverse films they are eager to see. According to this point, recommender systems should notice to this behavioral features of users, to recommend them an appropriate list of item to satidfy them. In another word, it will improve performance of recommender systems if they be able to suggest a list with high intradiversity to users who are fond of high diversity, and low diverse list to users who like it. Also, we proposed a hybrid model with adjustable level of personal diversity and accuracy. We used a user based collaborative filtering model and added a new phase to this model. Added phase consist of calculating similarity f user, based on how diverse they were in their past rates to different items. In this phase we clustered items in four classes to improve our model and also reduce the complexity of model. At last of this phase we take a similarity matrix and combine it with similarity matrix of user based collaborative filtering model. Our expriments on our proposed model showed that our model not only improve the personalized diversity of recommended list but also it reduced the MAE and RMSE errors. Since user based collaborative filtering is a commercial model and also model does not increase the complexity of model, our proposed model can be use in commercial application and online websites
  9. Keywords:
  10. Recommender System ; Social Networks ; Evaluation Metrics ; Collaborative Filtering ; Diversity Theory ; Personalized Diversity

 Digital Object List

 Bookmark

No TOC